Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

2x, not 10x: coding with LLMs in 2026

obryant.dev

Post reply on HN
Print view View original

191–200 of 260 posts

Re: 2x, not 10x: coding with LLMs in 2026

#191

Post by t-writescode »

metalspot wrote: post #163
steveBK123 wrote: post #8

Earlier quoted context omitted.

The other hard to measure part is that I see/hear a lot of LLM usage going towards dev work that was never prioritized before. Suddenly devs who were cranking out features with no interest in infrastructure are attacking giant refactors to make the code more understandable to the LLM. Other devs are using LLMs to build themselves quality of life SDLC tools completely separate from the core code base. Plenty of other…

llms are 100X for me, but only 10X goes into actual shipping code. the rest is all on doing the dev process the way it actually should be done: incremental POCs, refining specifications, modularization, exhaustive test suites with 95%+ logical unit test coverage, fuzzing unit test coverage, exhaustive e2e testing, with a fuzzer harness over e2e tests for long term simulation and scale testing, bechmark gated self imp…

It scares me how much of this is true - people truly weren't doing all of this before; but like ... some of us (like myself) were, and we were trying to push other people to write good code like that - code that helped everyone else's future velocity be faster - including our own on the next MR and the next one after that.

I'm glad you're doing all those right things; but, really, some of us were already doing most of, if not all of that. And some of us (like myself) were doing it very quickly, too, since we'd been doing it that way over a decade and constantly getting faster.

Re: 2x, not 10x: coding with LLMs in 2026

#192

Post by chr15m »

throwuxiytayq wrote: post #160
bigbuppo wrote: post #153

There's someone I know that jumped head-first into AI... they have embraced everything agentic. While setting up mail on their new iphone they encountered a problem. They had claude go through and do a bunch of tests against the server... allegedly. It didn't do any actual testing, though it did generate a long-winded report of what it pretended to try, fully blaming the server the whole time. Actual problem: he type…

Not sure what your point is, could you clarify? Are you saying that your friend is stupid? Or do you mean Claude is stupid? Both?

Claude lies, like all LLMs today.

Re: 2x, not 10x: coding with LLMs in 2026

#193

Post by rsoto2 »

switchbak wrote: post #103
pantelisk wrote: post #55

While I agree with the premise, I think this angle only applies on work one was going to do no matter what. The real power of these tools is that there are so many ideas people would like to try, but never have the time or motivation to pursue. So the comparison is not only "built with and without LLM" but "would you even build this if you didn't have the LLM?". The gap in productivity in this case is much more wide.

In this last week, I've performed experiments in an almost fully automated way - parallel to my main work, and easy to orchestrate. This has led to 3 parallel pieces of adjacent work that each speed up our build by quite a drastic margin. When combined, this is a massive improvement. None of this would have happened in the old days, as the research itself takes a long time to babysit and a lot of options to check. So…

or the AI could lead you to a completely unusable experiment and lead you to believe it was successful wasting weeks of work. Not saying it will happen to you, but there is a huge "survivorship" bias already in software for what tools are useful or not. Maybe it's the person and not the tools.

Re: 2x, not 10x: coding with LLMs in 2026

#194

Post by rsoto2 »

switchbak wrote: post #103
pantelisk wrote: post #55

While I agree with the premise, I think this angle only applies on work one was going to do no matter what. The real power of these tools is that there are so many ideas people would like to try, but never have the time or motivation to pursue. So the comparison is not only "built with and without LLM" but "would you even build this if you didn't have the LLM?". The gap in productivity in this case is much more wide.

In this last week, I've performed experiments in an almost fully automated way - parallel to my main work, and easy to orchestrate. This has led to 3 parallel pieces of adjacent work that each speed up our build by quite a drastic margin. When combined, this is a massive improvement. None of this would have happened in the old days, as the research itself takes a long time to babysit and a lot of options to check. So…

Figured out a better way of explaining it:

dev A knows exactly what the program should do and how to verify the AI output

dev B thinks they know what they are doing but are actually misguided by bad psycophantic AI output they have incorrectly verified.

both work on product C

Re: 2x, not 10x: coding with LLMs in 2026

#196

Post by xbmcuser »

I am not a programmer but for me the return is 100x as I am automating stuff that I knew could be automated but did not know how to but at the same time it was not worth the effort to write it down and pay someone to code it for me. But now I have created so many python and bash scripts that save me hours and hours a month. In my opinion the people that can most take advantage of the programing capabilities are people that least use it for such or know the how they could automate a lot of the repetitive tasks they have to do at work or on their own time.

Re: 2x, not 10x: coding with LLMs in 2026

#197

Post by robomartin »

Well, it's interesting. I've been using LLM's for coding for over two years. My findings are somewhat consistent: Feed it jobs in small chunks and it can be a great auto-complete, beyond that and it turns into a mess very quickly.

I'm two weeks into a project that has been interesting and also has confirmed the above yet again. I am porting a Windows application written in Python to C# using Avalonia UI.

I have never used C# or Avalonia UI. I've written Windows applications in other languages, never C#. I don't know the language, libraries, etc.

At first I told Codex: Here's the source code, port it. I just had to run that test.

Well, it didn't end well. I'll describe it as a frustrating set of prompts that seemed to result in the implementation going in circles with constant problems being introduced, breaking-fixing-breaking, etc.

I then started again with a clean repository and played the role of the architect with full documentation in the form of code. File-by-file, I had it port modules to effectively develop an operational foundation for the classes, methods, properties, abstractions, hardware interfaces, etc. in the original Python program. That went well, was very fast and a good experience. I am running Codex in JetBrains Rider and the integration is excellent. The native OpenAI Codex application is an absolute dog...it pegs all my cores at 100% while doing nothing.

Once all the underlying infrastructure was ported and, to the extent possible, individually tested, I threw UI integration at it. This happened quickly and OK from it's-ugly-but-I-can-use-it perspective. Codex seems to be way out of its element when it comes to UI/UX understanding. Funny examples like placing a button on top of an image with the image covering the button because it had a higher z order. Or finally placing the button at the correct z order but making it transparent on-hover. Funny stuff when you are moving slowly and you see it happen. It just proves that there is no understanding whatsoever.

With all of that and the experimentation, I'll estimate that a six month project will be cut down to three to four weeks. Another month and it will probably be a much better program with new features and more advanced capabilities.

And I have not touched a single line of code. Developing solid prompts is the absolute key, something that you can only really learn by doing and through lots of experimentation.

Yes, of course, I fed it working code. I think it could be very different if I were to start something from a blank slate.

Re: 2x, not 10x: coding with LLMs in 2026

#198

Post by rayiner »

coderatlarge wrote: post #150
rayiner wrote: post #108

Earlier quoted context omitted.

> There were so many things that no one will ever study and won’t give humanity any benefit but I use everyday to make my life better Agreed. I haven’t been this excited by computers since I got broadband DSL in 1998.

i consider it as a very specialized subset (subgenre) of video games for a very narrow audience

You know how non-coders basically build crazy business apps in Excel? LLMs, even if they don't get any better, seem at least 10x more useful than Excel. Why do you say it's for a narrow audience?

Re: 2x, not 10x: coding with LLMs in 2026

#199

Post by wwind123 »

I think it'd take some time for people to figure out what kind of harness or workflow would work best with each individual codebase or team culture. AI certainly can help speed things up a lot, but for now the humans driving the AI need discipline to follow good software-engineering practice, or at least have multiple AI agents critique each other's work, instead of just pushing out whatever one AI agent writes.

Re: 2x, not 10x: coding with LLMs in 2026

#200

Post by jatora »

nathell wrote: post #82
pantelisk wrote: post #55

While I agree with the premise, I think this angle only applies on work one was going to do no matter what. The real power of these tools is that there are so many ideas people would like to try, but never have the time or motivation to pursue. So the comparison is not only "built with and without LLM" but "would you even build this if you didn't have the LLM?". The gap in productivity in this case is much more wide.

I think the ideas worth pursuing still require substantial time and motivation, even with LLMs. Otherwise you just end up with a half-baked sloppy artifact that nobody will ever use or study, for a net negative benefit to humanity.

People need to abandon the notion of every tool you build needing to fit the market. This is not the direction of software anymore when developers can all build tools that are bespoke to their own workflows and needs.
Post reply on HN